月間参加
月額私のコンテンツの定期的な席。購読者はキューに入るか、枠よりも購読者が多い場合はランダム抽選で、ビデオ本体、クレジット、または説明に登場します。
月額
価格はお問い合わせ
次に再生
But how do AI images and videos actually work? | Guest video by Welch Labs 38 分 
I Investigated India’s Biggest Smartphone Controversy 42 分 
Игра Без Сохранений 2 (Мото VLOG) 58 分 
Sam Altman Shows Me GPT 5... And What's Next 66 分 
Google Pixel 11 Hands on - The Phone Crisis Begins. 14 分 
SEXTANT 2 分 
UNDEAD DOGS 1 分 
I hit rock bottom 17 分 
I Tested YouTubers' Favourite Products 30 分 
Buying a 1 euro House in Italy - is it a scam? 42 分 
I'm Actually Short 1 分 
Fixing the Most Dangerous Dam in the World 23 分 
Seas #TeamSeas 3 分 
India’s internet war 9 分 
EXTREME HIDE & SEEK IN FOOTBALL STADIUM 46 分 
Spreading Poop Everywhere Situation is Crazy 12 分 
How the Iran War Spiked Oil Prices 18 分 
What toys have kids played with throughout history? 6 分 
How Taiwan is Preparing to Get Invaded 17 分 
Hurricane vs. Tiny Houses 23 分 
How Earthquake Bearings Work 19 分 
Name That Infrastructure (Ep. 10) #shorts 1 分 
Practical Magic (1998) - The Exorcism of Nicole Kidman! (10/11) | Movieclips 6 分 
Name That Infrastructure (Ep. 13) #shorts 1 分 
They Call It A "Drowning Machine" | Fascinating Horror Shorts 1 分 
Stopping the Unstoppable 16 分 
The Hidden Engineering of Niagara Falls 16 分 
China’s Plan to Block All Submarines 21 分 
There’s something wrong with Hallmark’s youtube channel 19 分 
How Disney Built A Realistic Lightsaber 17 分 
How to undo the damage of sitting all day | The Gray Area 42 分 
Why straight dating is harder than ever | The Gray Area 56 分 
Why avalanches are most deadly when they stop - Simon Trautman 6 分 
My Favorite Twilight Knockoff 101 分 
How We Built the ISS - Part 2 29 分 
How Self Storage Consumed America 19 分 
How has technology influenced literature?: Crash Course Latin American Literature #11 11 分 
The Failed Logistics of Hurricane Katrina 21 分 
How did ocean invertebrates end up on Mt. Everest?! 1 分 
What Happens When Science Clashes with the Public?: Crash Course Scientific Thinking #7 13 分 
How Private Equity Ruined American Youth Sports 19 分 
Why Fast Food Got So Expensive 20 分 
Name That Infrastructure (Ep. 3) #shorts 1 分 
You've Never Seen A Wheelchair Like This 17 分 
What’s Ruining Our Ruins? 12 分 
Do Penguins Control the Weather? 8 分 
Peppa Pig Tales 2026 🤔 Can Peppa Find Out if BABY EVIE Has a TWIN?! 🍼 BRAND NEW Peppa Pig Episodes 121 分 
Peppa's OUTDOOR Cinema! #PeppaPig #Shorts 1 分 
Pizza in Italy 🍕 #shorts #peppapig 1 分 
I almost quit YouTube.... 24 分 
These Rocks Are Older Than the Sun 13 分 But how do AI images and videos actually work? | Guest video by Welch Labs
2025 38 分
Diffusion models, CLIP, and the math of turning text into images Welch Labs Book: https://www.welchlabs.com/resources/imaginary-numbers-book Sections 0:00 - Intro 3:37 - CLIP 6:25 - Shared Embedding Space 8:16 - Diffusion Models & DDPM 11:44 - Learning Vector Fields 22:00 - DDIM 25:25 - Dall E 2 26:37 - Conditioning 30:02 - Guidance 33:39 - Negative Prompts 34:27 - Outro 35:32 - About guest videos Special Thanks to: Jonathan Ho - Jonathan is the Author of the DDPM paper and the Classifier Free Guidance Paper. https://arxiv.org/pdf/2006.11239 https://arxiv.org/pdf/2207.12598 Preetum Nakkiran - Preetum has an excellent introductory diffusion tutorial: https://arxiv.org/pdf/2406.08929 Chenyang Yuan - Many of the animations in this video were implemented using manim and Chenyang’s smalldiffusion library: https://github.com/yuanchenyang/smalldiffusion Cheyang also has a terrific tutorial and MIT course on diffusion models https://www.chenyang.co/diffusion.html https://www.practical-diffusion.org/ Other References All of Sander Dieleman’s diffusion blog posts are fantastic: https://sander.ai/ CLIP Paper: https://arxiv.org/pdf/2103.00020 DDIM Paper: https://arxiv.org/pdf/2010.02502 Score-Based Generative Modeling: https://arxiv.org/pdf/2011.13456 Wan2.1: https://github.com/Wan-Video/Wan2.1 Stable Diffusion: https://huggingface.co/stabilityai/stable-diffusion-2 Midjourney: https://www.midjourney.com/ Veo: https://deepmind.google/models/veo/ DallE 2 paper: https://cdn.openai.com/papers/dall-e-2.pdf Code for this video: https://github.com/stephencwelch/manim_videos/tree/master/_2025/sora Written by: Stephen Welch, with very helpful feedback from Grant Sanderson Produced by: Stephen Welch, Sam Baskin, and Pranav Gundu Technical Notes The noise videos in the opening have been passed through a VAE (actually, diffusion process happens in a compressed “latent” space), which acts very much like a video compressor - this is why the noise videos don’t look like pure salt and pepper. 6:15 CLIP: Although directly minimizing cosine similarity would push our vectors 180 degrees apart on a single batch, overall in practice, we need CLIP to maximize the uniformity of concepts over the hypersphere it's operating on. For this reason, we animated these vectors as orthogonal-ish. See: https://proceedings.mlr.press/v119/wang20k/wang20k.pdf Per Chenyang Yuan: at 10:15, the blurry image that results when removing random noise in DDPM is probably due to a mismatch in noise levels when calling the denoiser. When the denoiser is called on x_{t-1} during DDPM sampling, it is expected to have a certain noise level (let's call it sigma_{t-1}). If you generate x_{t-1} from x_t without adding noise, then the noise present in x_{t-1} is always smaller than sigma_{t-1}. This causes the denoiser to remove too much noise, thus pointing towards the mean of the dataset. The text conditioning input to stable diffusion is not the 512-dim text embedding vector, but the output of the layer before that, [with dimension 77x512](https://stackoverflow.com/a/79243065) For the vectors at 31:40 - Some implementations use f(x, t, cat) + alpha(f(x, t, cat) - f(x, t)), and some that do f(x, t) + alpha(f(x, t, cat) - f(x, t)), where an alpha value of 1 corresponds to no guidance. I chose the second format here to keep things simpler. At 30:30, the unconditional t=1 vector field looks a bit different from what it did at the 17:15 mark. This is the result of different models trained for different parts of the video, and likely a result of different random initializations. Premium Beat Music ID: EEDYZ3FP44YX8OWT
このクリエイターと仕事をする
このビデオのクリエイターが公開したパッケージ。
このクリエイターはまだ料金を設定していません。リクエストを送信してください — 何かが請求される前に価格が合意されます。
保証配置 — 新規ビデオ
一回限り次のビデオに10秒以上の保証出演。価格は挿入が作品のどこに配置されるかによります。
- 最初の1～3分
- ビデオの中盤
- ビデオの終盤
価格はお問い合わせ
- 最初の1～3分
- ビデオの中盤
- ビデオの終盤
保証配置 — ショート
一回限り既存のビデオからカットしたショート動画内に10秒以上の保証出演。
一回限り
価格はお問い合わせ